Mid way through the announcement season, profits in the US, UK and Europe are coming in stronger than expected, and the momentum in analysts’ expectations has improved. The FTSE 100, buoyed by a pick-up in commodities and the currency impact post-Brexit has actually seen increases in analysts’ forecasts.
This picture is relatively encouraging; the average value for this measure for the last thirty years has been close to -2% for both the UK and US, so net upgrades in the UK are quite unusual. It is also the case that the MSCI AC World spends a significant amount of time with negative revisions to profits expectations. The phenomenon of analyst over-optimism in profits forecasts is well documented.
These moves correlate reasonably well with an improvement in economic momentum in both the US and Europe in recent months, and it is also the case that improving profits momentum is not only a result of a recovery in commodities. The three largest sectors in the MSCI AC World: Financials, IT and Consumer Discretionary (which together make up 45% of the index, relative to energy at around 7%) have also seen a modest improvement in momentum from their lows earlier in the year.
Why should we care?
It may seem odd that we should pay attention to the predictions of the analyst community, given our own scepticism with regards to forecasting and forecasters. It is almost certainly true that the charts above tell us far more about what has happened than what will happen.
However, assessing consensus profits expectations can tell you something about the nature of market optimism or pessimism that may not be captured in valuation signals alone. One study looking at country selection between 1987 and 1991 and another looking at stock selection in Europe between 1987 and 1999 suggested that positive profits revisions had been an indicator of future outperformance.
From our perspective there is a behavioural explanation for these findings. The idea is that analysts react slowly to profits news, which itself tends to display trending properties. And both the studies look at time periods when there were strong positive trends which differed from past experience; the first study captures the red circle in figure 4 below, and the second captures the long period in 1990s marked in green (we have used the S&P here due to the longer history available).
There is therefore no reason that the phenomenon of earnings revisions leading stock prices should persist. It is simply a function of a trending environment which marks a structural shift that investors are slow to pick up on.
In the 1990s, analysts and investors were slow to realise that they had moved into a new regime of more pronounced profits growth. Ultimately they came to extrapolate that regime, resulting in the tech bubble of 1999/2000. What do persist however are the more general types of behavioural bias captured by these studies, such as anchoring and availability bias.
So what about today? It is no surprise with no major trend in place (or a transition to slower earnings growth if you believe some commentators); earnings revisions have not been useful in telling us about prices. If anything it looks like the relationship may even be the other way round: prices have ‘forecasted’ downward profits revisions.
This is also true at a global level, where the strongest relationship over the recent past is for price moves to lead subsequent revisions in profits expectations (the relationship is strongest with a nine month lag, for those that are interested).
Again this is a function of the environment. The financial crisis plays a disproportionate role on the sample period, while profits in general have disappointed in recent years. Investors must consider whether this environment is likely to persist before judging whether this relationship is of any use.
Profits are the driver of equity returns over the long term, and so it is essential for valuation that the structural dynamic is taken into account. At the same time, there are important lessons to be gleaned from shorter term changes to profits expectations. To understand these though, we cannot rely on searching for patterns and using regressions, but instead must seek to assess whether the broader behavioural biases illustrated in past periods will be applicable in the future and how they will manifest themselves.